Dplyr

Quantitative Methodology (UPF)

Jordi Mas Elias

https://www.jordimas.cat/

Summary

Pipe
Row dplyr functions
Column dplyr functions
Transform variables

Warm up

R learning curve

RStudio workflow

Load packages.

library(dplyr)
library(ggplot2)
library(stringr)

Pipe

One function

f(d$v)
f(d, v)
d |> f(v)
1:100 |> 
   sample(4)

Various functions

f1(f1(d, v1), v2)
d |> 
  f1(v1) |> 
  f2(v2)

Pipe

Without pipe

go_to_upf(take_transport(get_dressed(get_out_of_bed(wake_up(me,
time = "8:30"), side = "left"), pants = TRUE, shirt = TRUE), 
bus = TRUE, metro = FALSE), mqi = FALSE, bar = TRUE)

With pipe

me |> 
  wake_up(time = "8:30") |> 
  get_out_of_bed(side = "left") |> 
  get_dressed(pants = TRUE, shirt = TRUE) |> 
  take_transport(bus = TRUE, metro = FALSE) |> 
  go_to_upf(mqi = FALSE, bar = TRUE)

Example taken from here and here.

Row dplyr functions

Filter

Reduces the number of rows according to a certain criteria

df |> 
  filter(logic_vector)

Always is the result of a logic vector.
Can use & (AND), | (OR), ! (NOR).

Arrange

Arranges from less to more…

df |> 
  arrange(vector)

…or from more to less.

df |> 
  arrange(desc(vector))

Count

Counts the number of categories in a vector¹…

df |> 
  count(vector)

…and orders the results.

df |> 
  count(vector, sort = T)

Column dplyr functions

Select

Selects vectors.

df |> 
  select(vector1, vector4, vector6:vector9)

Removes vectors.

df |> 
  select(-vector2)

Rename

Renames the vector.

df |> 
  rename(new_name = Old.Vector.Name)

Mutate

Modifies the values of a vector¹…

df |> 
  mutate(vector5 = vector5 * 100)

… or creates a new vector².

df |> 
  mutate(new_vector = vector5 + vector6)

Several operations can be combined.

df |> 
  mutate(vector5 = vector5 * 100,
         new_vector = vector5 + vector6)

Summarize

Summarizes data.

df |> 
  summarize(name = sum(vector))

Different elements can be summarized:

df |> 
  summarize(name1 = sum(vector),
            name2 = mean(vector),
            n = n())

*Mind the argument na.rm = T.

Group_by

Always combined with another function (e.g. summarize, filter, mutate), it groups the data by the values of a vector¹…

df |> 
  group_by(vector) |> 
  summarize(name1 = sum(vector))

*With group_by and summarize, we change the unit of observation of the dataset.

Recoding vectors

Recoding

When we recode variables (vectors), we lose information.

Destí	Funció
Binària	`if_else()`
Categòrica	`case_when()`
Ordinal	`factor()`
Qualsevol	`recode()`
Altres	`as.numeric()`, `as.character()`, `as.Date()`, etc.

Boolean operators

AND (&): TRUE if all conditions are met.
OR (|): TRUE if any condition is met.
NOT (!): TRUE if conditions are not met.

If_else

To a dichotomous / binary / dummy variable.

df |> 
  mutate(new_name = if_else(logic operation, true, false))

Case_when

case_when(logic operation ~ "C1"
          logic operation ~ "C2",
          logic operation ~ "C3",
          ...,
          TRUE ~ "CN")

Factor

df |> 
  mutate(new_vector = factor(wb$income_group, 
                             ordered = TRUE,
                             [levels o labels = ...]))

Recode

df |> 
  mutate(new_vector = recode(vector, 
                             old_value = "new_value"))

As functions

as.numeric(vector)
as.factor(vector)
as.character(vector)
as.integer(vector)
as.Date(vector)